There are a lot of interesting advanced projects (AI, automated driving, etc.), but I would be worried about the search core right now as CEO.
The proliferation of content-mills and SEO companies is really starting to take a toll on search results. I find myself avoiding Google and going to niche sources directly now (like Wikipedia or StackOverflow) just to avoid the Demand Media trash that is everywhere now.
I think Google should really invest in a new approach to rank -- one that could maybe focus on the content itself using AI and NLP instead of link juice.
I agree completely about search quality, as my comment history here and elsewhere might indicate. The majority of my Google searches anymore return very low quality results without some serious query mangling or heavy filtering. At this point I too skip Google completely as much as possible and go straight to the "niche sources".
People like Matt Cutts at Google unfortunately don't seem to care about the extent low-quality sources have gamed Google - Cutts is still even defending Mahalo as far as I know.
Also, I forgot to mention the amount of outdated forum posts that you get with technical searches - that's the other huge problem with search quality for me aside from content mills and content scrapers. If I go directly to Stack Overflow et al to do a search, I can at least be reasonably sure to get an up-to-date answer.
More recent forum results should definitely be weighted higher than older ones, for a start.
If you're not happy with the dates of the results we show now, one option is to click on "More search tools" on the left-hand bar on the search result page. Then pick a date option like "Past week" or "Past month" or "Past year."
I don't feel like I should have to narrow down date ranges, though I end up doing so every time a search returns forum posts. The algorithm ought to know that something posted six weeks ago is much more likely to be relevant than something posted six years ago, but older posts seem to instead have more weight in the unfiltered results rather than less.
google's date features (as matt has point out) are one of the reasons i haven't left google for bing completely. sure, getting forum post search results about coding an API from 2006 are usually useless, but in bing i can't even say "within the last month." google wins there, currently.
cdr, we've taken action on Mahalo in the past when we've discovered practices that we would consider spammy. At the same time, given the impact that Google has on the web it's important that we not take action on a site just because we like or dislike it. We try to make sure that we only take action on sites when they violate our quality guidelines.
My issue is the increasingly low quality of Google search results. The problem is not with just one particular site, as distasteful as I find Mahalo, but a class of sites. If you don't consider the SEO practices of the content mills that are infesting the majority of Google results "spammy", IMO you should either reconsider what practices you consider spammy or refine the search algorithm to produce higher quality results.
I recall reading a thesis linked on HN that the only technologically knowledgeable users care about search quality - that the average, non-technical google user doesn't care that a result is from low quality content mill, only that they get some sort of information that sort of seems relevant. I think (hope?) that technical users are the leading edge here, and google's reluctance to act on content mills and improve search quality will end up hurting.
Hi Matt. I would love an option to blacklist domains from my results. That gets around the issue of Google picking winners and losers and would make pickiest 1% of users (like me) very happy.
Even a modicum of human intervention would help far more than any possible investment into more automated ways to address this problem would. At some point, Google just needs to gird their loins and bring the banhammer down. Yeah, that has problems. I could write about them at length. But that's not the question; the question is now which approach has more problems. I'm willing to try the banhammer at this point. (Though an accompanying investment into Real People that we could actually reach in case of error would be nice, too.)
Google is definitely vulnerable in terms of search result quality.
They have tremendous market share, but search result quality has been stagnant or even declining for a while now.
Page previews are a negative for me, they're distracting and annoying, and every time one comes up I inevitably think "I wish they'd spent this time improving the results".
that would require a fundamental paradigm shift from what they currently use to determine the ranking order of search results. Backlinks as a ranking factor will remain a core feature for the indefinite future I would imagine.
I would say they will continue to pursue social/local ranking algorithms. What will most likely happen is that the idea of an objective "high ranking" won't really exist in the future, as all rankings will be personalized to your own (and social network's) habits/interests/history. Combine that with their pursuit of increased speed everywhere, instant everything, and live updates, and you'll be seeing a very different kind of search engine in even just a few years.
Comments
There are a lot of interesting advanced projects (AI, automated driving, etc.), but I would be worried about the search core right now as CEO.
The proliferation of content-mills and SEO companies is really starting to take a toll on search results. I find myself avoiding Google and going to niche sources directly now (like Wikipedia or StackOverflow) just to avoid the Demand Media trash that is everywhere now.
I think Google should really invest in a new approach to rank -- one that could maybe focus on the content itself using AI and NLP instead of link juice.
I agree completely about search quality, as my comment history here and elsewhere might indicate. The majority of my Google searches anymore return very low quality results without some serious query mangling or heavy filtering. At this point I too skip Google completely as much as possible and go straight to the "niche sources".
People like Matt Cutts at Google unfortunately don't seem to care about the extent low-quality sources have gamed Google - Cutts is still even defending Mahalo as far as I know.
Also, I forgot to mention the amount of outdated forum posts that you get with technical searches - that's the other huge problem with search quality for me aside from content mills and content scrapers. If I go directly to Stack Overflow et al to do a search, I can at least be reasonably sure to get an up-to-date answer.
More recent forum results should definitely be weighted higher than older ones, for a start.
If you're not happy with the dates of the results we show now, one option is to click on "More search tools" on the left-hand bar on the search result page. Then pick a date option like "Past week" or "Past month" or "Past year."
I don't feel like I should have to narrow down date ranges, though I end up doing so every time a search returns forum posts. The algorithm ought to know that something posted six weeks ago is much more likely to be relevant than something posted six years ago, but older posts seem to instead have more weight in the unfiltered results rather than less.
The point of the search engine, at least to me as a user of it, is to get results that are relevant to my needs.
Every tweak you require me to apply in order to get the results I want is an area in which someone else can beat you.
google's date features (as matt has point out) are one of the reasons i haven't left google for bing completely. sure, getting forum post search results about coding an API from 2006 are usually useless, but in bing i can't even say "within the last month." google wins there, currently.
cdr, we've taken action on Mahalo in the past when we've discovered practices that we would consider spammy. At the same time, given the impact that Google has on the web it's important that we not take action on a site just because we like or dislike it. We try to make sure that we only take action on sites when they violate our quality guidelines.
My issue is the increasingly low quality of Google search results. The problem is not with just one particular site, as distasteful as I find Mahalo, but a class of sites. If you don't consider the SEO practices of the content mills that are infesting the majority of Google results "spammy", IMO you should either reconsider what practices you consider spammy or refine the search algorithm to produce higher quality results.
I recall reading a thesis linked on HN that the only technologically knowledgeable users care about search quality - that the average, non-technical google user doesn't care that a result is from low quality content mill, only that they get some sort of information that sort of seems relevant. I think (hope?) that technical users are the leading edge here, and google's reluctance to act on content mills and improve search quality will end up hurting.
I appreciate the direct response, though.
Hi Matt. I would love an option to blacklist domains from my results. That gets around the issue of Google picking winners and losers and would make pickiest 1% of users (like me) very happy.
Wouldn't you be able to do that with Blekko?
Even a modicum of human intervention would help far more than any possible investment into more automated ways to address this problem would. At some point, Google just needs to gird their loins and bring the banhammer down. Yeah, that has problems. I could write about them at length. But that's not the question; the question is now which approach has more problems. I'm willing to try the banhammer at this point. (Though an accompanying investment into Real People that we could actually reach in case of error would be nice, too.)
You mean like Google Quality Raters?
Google is definitely vulnerable in terms of search result quality.
They have tremendous market share, but search result quality has been stagnant or even declining for a while now.
Page previews are a negative for me, they're distracting and annoying, and every time one comes up I inevitably think "I wish they'd spent this time improving the results".
that would require a fundamental paradigm shift from what they currently use to determine the ranking order of search results. Backlinks as a ranking factor will remain a core feature for the indefinite future I would imagine.
I would say they will continue to pursue social/local ranking algorithms. What will most likely happen is that the idea of an objective "high ranking" won't really exist in the future, as all rankings will be personalized to your own (and social network's) habits/interests/history. Combine that with their pursuit of increased speed everywhere, instant everything, and live updates, and you'll be seeing a very different kind of search engine in even just a few years.